Division of Computing Science and Mathematics University of Stirling

Stereo audio stimuli.

Recorded using a flat panel with 2 microphones (see L.S. Smith, S. Collins, Determining ITDs using two microphones on a flat panel during onset intervals with a biologically inspired spike based technique IEEE Transactions on Audio, Speech and Language Processing, 15, 8 (in press), 2007).

Recording Info

The microphones (AKG C417) are mounted flush at the top left and right (406 mm apart) of a plywood panel of size 440 by 330 mm, from which they are acoustically insulated. This is placed on a small table, approximating the possible location of microphones on a flat panel display. Sounds are played through Creative Soundworks Creative SB35 speakers: these are not omnidirectional, and the speakers were always pointed towards the midline of the two microphones. Sound was played and recorded either using an Marian Marc 2 soundcard, or, later, a MOTU828 Mark 2 at 96 KSamples/second, 16 bits linear. MATLAB was used for numerical calculations. The room is 3220 by 2558 by 2726 mm high (the ceiling is crenelated, with maximum height 3276 mm). The longer room walls are of painted breeze-block. One of the shorter walls is almost entirely window, and the other made of plasterboard with a large whiteboard and a door on it, so that the walls are all almost flat surfaces. The reverberation of the room may be altered by adjusting a curtain which can cover three of the four walls. The floor is covered in a thin nylon carpet, and the ceiling is made of concrete. The room is highly reverberant when the curtains are entirely open. With the curtains open, the $T_{60}$ (time for sound level to drop 60 dB) is approximately 320 ms, and for the curtain closed condition, it is approximately 160 ms. We have also carried out experiments with the microphones at the same distance apart, and the same height from the table, separated by air. Very similar results are obtained.

Stimuli Info

Directory pinkpip:

A brief wideband noise pulse was played from a distance of 1311 mm, in the same plane as the microphones, at angles of -30 degrees to +90 degrees in 10 degrees intervals. Sounds were recorded with the curtains entirely closed. (Results with the curtains open are almost identical.) The sound input to the loudspeaker consisted of pink noise (i.e. equal energy in each octave) with a 1 ms attack time, lasting 180 ms in all.

Directory SpeechWithNoise:

Contains 6 directories, two for 1khz background noise, 2 for pink noise background noise, and two for speech with delayed speech. Each pair has one recorded with the curtains open (_wall), and one recorded with the curtains closed (_curtains). The first four are based on a brief period of male speech (one author, LSS, counting from one to five), with other sound sources (a 1 kHz tone and some pink noise) at different angles, while varying the SNR (measured peak energy to peak energy). The file names are of the form



N - noise
sign = + or - for left or right
angle = 0 or 30 degrees
S - speech
sign = + or -
angle = 0 or 30 degrees
SNR peak-peak measurement


The last two have filenames of the form


spxxxspyyy.au


where xxx is one angle (-30degrees) and yyy is the other angle (0 or +30 degrees. The sound levels are the same for both sides.

Directory malefemale:

These all have the male speaker at 0 degrees, and the female at 30 degrees, with different SNRs.

The filke names are

p-00_mfxxxvoice.au

where xxx is the SNR (-00 is 0dB).

Directory moving

One signal, LSS speaking the words "The two thousand and six international conference on acoustics speech and signal processing". whilst walking round the front of the setup.


For further information, contact

Leslie Smith, email lss at cs dot stir dot ac dot uk



Back to Professor Smith's Research page.

Last updated: Thursday, 16-Aug-2007 10:30:04 BST

If you have any difficulties accessing this page, or you have any queries/suggestions arising from this page, please email:
Prof Leslie S Smith (lss(nospam_please)@cs.stir.ac.uk)

computing logos